Can you post a small working example with some dummy data that includes the PyMC model? I have a suspicion of what is wrong – you’re mixing compiled aesara functions (aef) and symbolic tensors (generated by aesara_code). You can inject pre-defined PyMC variables into sympy code using a cache dictionary and the aesara_code function, but it’s not totally clear to me how t should be defined so I can’t give a concrete suggestion on how to implement it.